English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Learning with limited data is one of the biggest problems of machinelearning. Current approaches to this issue consist in learning generalrepresentations from huge amounts of data before fine-tuning the model on asmall dataset of interest. While such technique, coined transfer learning, isvery effective in domains such as computer vision or natural langageprocessing, it does not yet solve common problems of deep learning such asmodel interpretability or the overall need for data. This thesis explores adifferent answer to the problem of learning expressive models in dataconstrained settings: instead of relying on big datasets to learn neuralnetworks, we will replace some modules by known functions reflecting thestructure of the data. Very often, these functions will be drawn from the richliterature of kernel methods. Indeed, many kernels can reflect the underlyingstructure of the data, thus sparing learning parameters to some extent. Ourapproach falls under the hood of "inductive biases", which can be defined ashypothesis on the data at hand restricting the space of models to exploreduring learning. We demonstrate the effectiveness of this approach in thecontext of sequences, such as sentences in natural language or proteinsequences, and graphs, such as molecules. We also highlight the relationshipbetween our work and recent advances in deep learning. Additionally, we studyconvex machine learning models. Here, rather than proposing new models, wewonder which proportion of the samples in a dataset is really needed to learn a"good" model. More precisely, we study the problem of safe sample screening,i.e, executing simple tests to discard uninformative samples from a dataseteven before fitting a machine learning model, without affecting the optimalmodel. Such techniques can be used to prune datasets or mine for rare samples.

On Inductive Biases for Machine Learning in Data Constrained Settings